1. Introduction

Metal nanostructures are subject of many studies since they exhibit collective charge density oscillations (surface plasmons), which are able to strongly enhance electromagnetic fields [1–3]. It is well known that the resonances of these oscillations depend on the particle material [4], the dielectric constant of the surrounding [5], as well as the particle geometry [6,7]. In the present study, we exploit the geometry dependence to tune the surface plasmon resonance (SPR) of rod-shaped nanoparticles with lengths in the micrometer range. These structures are of special interest for sensing applications in the infrared (IR) spectral region, e.g. for surface-enhanced IR spectroscopy (SEIRS) [8]. By changing the rod length, the spectral resonance position can be tuned [9–11] to match specific IR active vibrations of molecules. Resonant excitation leads to local electric field and absorption enhancement and, hence, the detection of characteristic, finger-print-like vibrations of molecules adsorbed to the nanoantennas can be substantially improved [12].

As conventional IR light sources (e.g. globars) have low intensity, SEIRS experiments are typically carried out on arrays of identical nanoantennas. Additional enhancement of SEIRS signals can be achieved by exploiting the extraordinary near-field enhancement of nanoantennas interacting with each other across very small gaps (nm range) between their tip ends [9,13]. Such nanogaps yield near-field amplitudes which are several orders of magnitude higher than those of uncoupled structures [9] and, hence, promise to drastically improve the sensitivity of sensing approaches. The controlled fabrication of nanoscale gaps is therefore highly desired, but their reliable preparation in a parallel process suited for large-scale arrays is a challenging task still hampering the success of SEIRS. For example, cutting nanowires with focused ion beams is a rather delicate problem since remaining material may bridge the gap as recently shown by IR spectroscopy [14]. Another important issue in SEIRS is tuning of the plasmonic resonance to the vibrational frequency of interest [12]. Thus, it is important to know how the nanogaps shift the resonances.

To address this problem without the use of FIB, we applied optically induced metal deposition to narrow gaps between IR-resonant nanorods manufactured by conventional electron beam lithography. Covering the nanorod arrays on the substrate with a solution of tetrachloroaureate (HAuCl₄) and illuminating them with visible light leads to the reduction of the gold salt, a site-selective gold deposition, and, thus, to a gradual growth of the gold nanostructures. We investigated in how far it is feasible to narrow gaps between nanoantennas of an array in a controlled way. To this end, we determined the alterations of the particle extinction spectra during subsequent growth steps and correlated them to geometrical changes imaged by scanning electron microscopy (SEM). With the obtained data, we analyzed sample processing and set-up and derived suggestions for further improvement of the photochemical method.

2. Materials and methods

2.1 Lithographic gold nanorod preparation

Gold nanorod arrays (overall size of 50 x 50 µm²) were fabricated by electron beam lithography at the MANA Foundry station at NIMS, Tsukuba, Japan, on natural oxide-covered silicon substrates as described in [15]. Within one array, cuboid-shaped rods with rectangular cross-sections (width w ≈120 nm and height h ≈100 nm) are arranged in parallel lines, separated by an approximately 40 nm wide gap between the tip ends [see Fig. 1(a) ]. Furthermore, the distance d_y between the individual lines was set to 5 µm to avoid interaction between rods of different lines [16].

 

Fig. 1 Scheme of the photochemical nanoantenna growth experiment. (a) Sample layout with several EBL-written nanoantenna arrays. The SEM inset shows individual nanorods separated by a 40 nm gap. (b) The sample is located on a movable xyz-table of the IR microscope and covered with droplet of HAuCl₄ solution. A specific array is illuminated by the microscope’s halogen lamp. (c) Consequently, Au ions are formed in the solution and diffuse to the nanorods where gold deposition takes place.

Download Full Size | PDF

2.2 Microscopic infrared spectroscopy

Spectroscopic far-field measurements were performed with a commercial IR microscope (Bruker Hyperion 1000, numerical aperture of 0.52, direct current heated SiC as light source) coupled to a Fourier transform IR spectrometer (Bruker Tensor 27 with a LN₂ cooled mercury-cadmium-telluride detector, optical path purged with dried air). IR transmittance measurements of the gold nanorods on the substrate were performed at normal incidence with IR light polarized parallel to the long rod axis. The obtained spectra were normalized to those of bare substrate areas taken at least 50 μm away from any rod to eliminate all background features. For all measurements, a circular aperture with a diameter of 33.3 µm in the focal plane of the IR microscope was used. The IR spectra were acquired with at least 300 scans at a resolution of 8 cm⁻¹ in the spectral range from 1000 to 7000 cm⁻¹ (λ = 10 µm to λ = 1.4 µm).

2.3 Photo-induced metal deposition

Photo-induced metal deposition was carried out in the following way: First, a droplet (volume of 20 μl) of 1 mM gold salt solution, produced by dissolving HAuCl₄ (ABCR, 99.9%) in a low-viscosity immersion oil (Cargille #1160), was applied on the sample. The immersion oil was chosen to avoid evaporation of the solvent during the experiment as well as to provide optimal light transmission during exposure.

The sample was mounted onto a movable xyz-table of the IR microscope and the deposition area was selected through the microscope eyepiece. It is important to note that only photons with a sufficiently high energy are able to activate the reduction process [17]. Thus, to avoid any deposition during the positioning process, a color-glass filter was used to remove the short-wavelength portion of the white light spectrum used for illumination (Nikon halogen lamp, 12V, 100W). After defining the deposition area, the filter was removed from the beam path and the entire array was illuminated with the white light for a certain period of time.

Due to the opaque substrate, the sample was illuminated from the top by the halogen lamp as shown in Fig. 1(b). As the light had to pass through the gold salt solution, no tight focusing on the substrate was possible with the IR microscopic set-up and, hence, whole arrays were illuminated at once. Details of the metal deposition process are illustrated in Fig. 1(c): Upon irradiation, gold ions are formed in the gold salt solution. However, the full reduction of the gold salt only occurs in the presence of a surface perturbation of the substrate which works as a seed for the deposition process. For this reason, partially reduced species diffuse within the solution until they find suitable deposition sites (see [18,19] for a detailed description of the chemical process). The latter are provided by the surface of the gold nanorods, where the finally generated gold atoms are deposited. Note the high spatial selectivity of the process as no gold precipitates on the silicon surface.

After deposition, the sample was rinsed with pure ethanol for one minute, reinstalled in the IR microscope and IR transmittance spectra were recorded. This procedure (referred to as one growth step) was iterated several times to follow the evolution of the optical properties. An in situ observation of the particle growth was not possible with the present transmittance measurement technique due to the strong light absorption of the solvent. This problem can be principally overcome by use of attenuated total reflection (ATR), as recently shown for wet-chemical manipulation of gold island films [20].

3. Results and discussion

Figure 2(a) shows a series of relative transmittance spectra of one specific gold nanorod array with initial rod length of 700 nm [see Fig. 2(b)]. Note that all spectra were recorded with IR light polarized parallel to the long rod axis. Measurements with perpendicularly polarized light are not shown, since the transverse plasmonic resonance of the investigated nanoantennas is not located in the IR spectral range [15,21]. Three consecutive growth steps were performed on this array [spectra (1) to (3)] with irradiation times indicated in Fig. 2(a). After that, another nanorod array, located approximately 150 µm away from the previously treated one, was exposed to the white light (illumination time of 5 minutes) and spectrum (4) was recorded to test the selectivity of the metal deposition process.

 

Fig. 2 (a) Evolution of the relative transmittance of a gold nanorod array during subsequent growth steps and respective resonance position indicated in the inset. Starting from reference spectrum (0), IR measurements were performed after every growth step (illumination times are indicated). Although the fourth illumination step was carried out on an array located approximately 150µm away, the optical properties still change due to diffusion of the gold ions. The features at around λ = 4.2 µm and λ = 8 µm originate from CO₂ absorptions in the optical beam path and from the excitation of a thin-film surface phonon-polariton in the silicon oxide layer [15], respectively. (b) SEM image of three rods in the array recorded before the acquisition of spectrum (0). The geometric dimensions are: L ≈700 nm, w ≈120 nm, d_x ≈40 nm, d_y = 5 µm. (c) SEM image taken after acquisition of spectrum (4). Conductive connections between the nanorods have formed, indicated by arrows pointing downwards. However, some gaps are still present and gap sizes below 10 nm can be seen (arrow pointing upwards). The width of the rods has increased to about 160 − 170 nm. Note that only very small regions of the arrays are shown in the SEM images (overall array size is 50 × 50 μm²).

Download Full Size | PDF

Starting from the measurement before metal deposition [reference spectrum (0) in Fig. 2(a)], the position of the plasmonic resonance shows the expected behavior: it shifts to longer wavelengths with increasing illumination time, since the gaps become narrower and interaction increases [9,16]. A plot of the resonance frequency vs. irradiation time [inset of Fig. 2(a)] shows a linear dependence which enables easy process control. Thus, a targeted tuning of the spectral properties of the nanoantenna array is possible. The narrowing of the gaps can clearly be seen in the SEM image [Fig. 2(c)]. Furthermore, as expected, no deposition of gold occurs on the silicon substrate. These results clearly demonstrate the proof-of-concept, i.e. the suitability of the photochemical approach to tune the gap size in nanoantenna arrays on opaque substrates. However, a closer look at the data also indicates some difficulties which we discuss now.

While the extinction slightly increases during the first two growth steps most likely due to increased particle volume, a decrease of extinction accompanied by strong broadening is observed after the third and the following growth step [spectrum (3) and (4)]. In principle, two effects contribute to the decrease in extinction: First, smaller gap size and hence increasing interaction between individual nanorods results in a reduced far-field signal [9,16]. Second, the merging of some rods (with length L) to more elongated structures (2L, 3L, …) will also diminish the intensity of the main antenna resonance since longer structures are resonant at longer wavelengths. Due to the linear behavior between antenna length and resonant wavelength in the IR [16,22], one would expect an emerging peak at twice the wavelength of the fundamental resonance. And in fact, we observed such peaks in IR measurements of other arrays with shorter initial rod length. In Fig. 2(a), only a slightly increased extinction can be seen above λ = 8 µm in spectrum (4). Note that the spectral change observed in spectrum (4) compared to spectrum (3) implies that the growth process still continued even though the array was not directly illuminated. We attribute this behavior to diffusion of gold ions from their point of origin to other places outside the illumination spot. Nevertheless, SEM images, taken after acquisition of spectrum (4), confirmed that conductive connections between the nanorods have formed, while other nanorods were still separated by narrow gaps below 10 nm [see Fig. 2(c)].

As the SEM images of Figs. 2(b) and 2(c) show, the surface properties of the nanorods drastically change after metal deposition. It seems that rather big gold grains attach to the cuboid-like and relatively smooth lithographic rods, leading to grainy particles. However, this effect does not impose any problems as crystallinity and surface defects, respectively, of gold nanoantennas were found to only marginally influence the far-field optical response [22,23]. Moreover, recent studies showed a small influence of the surface roughness on the optical properties of plasmonic particles [24]. Furthermore, direct experimental evidence of the negligible impact of surface roughness is given in Fig. 3 . Here, SEM images of typical nanorods of non-interacting arrays (d_x = d_y = 5µm) [16] are shown: an array illuminated for 2 min [Fig. 3(b)] on the one hand, and a lithographically prepared array, which was not photo-chemically manipulated, on the other hand [Fig. 3(c)]. And in fact, the relative transmittance spectra of these arrays with rods of similar geometric dimensions suggest no significant differences in the far-field optical performance [Fig. 3(a)]. However, the near-field distribution might considerably differ, since the photo-chemically formed nanoparticles feature many “hot spots”, i.e. sites with very sharp tips, leading to additional local field enhancement [25,26]. Consequently, the increased surface roughness could even improve the sensitivity of SEIRS applications.

 

Fig. 3 (a) Relative transmittance spectra of non-interacting nanorod arrays (d_x = d_y = 5 µm) consisting of rods of similar geometric dimensions but different surface structure. The same number of rods contributes to each spectrum. In the SEM images of (b) and (c), representative examples of “rough” and “smooth” nanorods are shown.

Download Full Size | PDF

From the experimental findings, two major problems can be pointed out. First, the problem of gold ion diffusion which is most likely caused by the top illumination geometry combined with the missing possibility of tightly focusing the visible light onto the nanoantennas with the IR microscopic set-up. Second, the inhomogeneous growth within an array of nanoantennas hampers a direct correlation between gap size and geometric structure of the rod. Due to these difficulties we conclude that in order to ensure better applicability of the photochemical method, single nanorod pairs on VIS-transparent substrates should be used. In contrast to particle arrays on opaque Si substrates, light can be focused onto single antenna pairs from the bottom with a narrower focus using a laser and an inverted optical microscope. By addressing single antenna pairs, the effects of uncontrolled antenna merging and growth will be reduced. Suitable deposition parameters (light intensity, gold salt concentration, and exposure time) should be derived with such single-antenna samples and afterwards be applied to antenna arrays which are finally used for actual SEIRS experiments.

4. Summary and conclusion

In this paper, we have shown that the photochemical metal deposition technique can be used to reduce the gap size between µm-sized gold nanorods. Monitoring this technique with IR micro-spectroscopy, we were able to observe a red-shift of the plasmon resonance peak indicating increased plasmonic coupling. Besides the proof-of-concept-experiment, we furthermore discussed the influence of the emerging surface roughness and difficulties of the technique arising from the arrangement of the nanoantennas in an array (as it is typically used for SEIRS) and the opaque substrate. Hence, in future work, we will investigate the targeted growth of single nanorod pairs on transparent substrates. These experiments will allow us to determine parameters for the photochemical metal deposition which can then be applied reliably on antenna arrays.